Hello all,
I run Varnish on a number of servers now, but one thing that is regularly bothering me is how keepalive checks by ip_vs skew the hit ratio on slow servers (e.g. at night). Is this something that has come up but that I've missed? I'd like to find a solution to this problem, but the source code is daunting :) I'd like to probe you guys on ideas on how to get better stats in my (not very uncommon) scenario. An idea that I've had is to provide a way to mark certain hits as *req.do_not_log (*in vcl_recv?) or *req.no_statistics*. This would prevent that connection from getting used in statistics. Places where client_req is incremented would have to skip that increment. Other places would not increment cache_hit, cache_hitpass or cache_miss. It would touch a lot of code, but when done is backwards compatible, because those who don't set the counter are fine. Another way is to count twice, one *client_req* and one * client_selective_req* or so. Any thoughts? Best, Allard Hoeve
_______________________________________________ varnish-dev mailing list [email protected] https://www.varnish-cache.org/lists/mailman/listinfo/varnish-dev
